Fraquency domain and rolling bearings

Projects General Guidance
1

I'm looking for an Arduino library able to give the presence in frequenquency domain of a characteristic frequency of a bearing failure.

There are some characteristic frequencies in a frequenquency domain when I analyze the vibrational signal of a failed bearing.

These frequenciese tell us that in the bearing there are some problems.

The signals are generated by rotation of rolling bearings in a elettric motor.

I use an accelerometer with an arduino.

This is a very interesting project:

Can I found this frequency?

2

A google search for "arduino fft library" gives several links with code examples.
Have you tried using any of them?

3

I will be frustrated by the coarse resolution of the frequency bins.

See this post:

http://forum.arduino.cc/index.php?topic=299461.0

4

sossio89:
I will be frustrated by the coarse resolution of the frequency bins.

See this post:

http://forum.arduino.cc/index.php?topic=299461.0

I guess that is why the authors of the report you first quoted chose to use Matlab for the analysis of the results which were collected from the accelerometer by the Arduino.

5

Maybe worth looking at the video on this page. Specifically 34:56 into the video where they discus/demo FFT capabilities of the chip.

6

sossio89:
I will be frustrated by the coarse resolution of the frequency bins.

Then use more bins and perhaps a more powerful processor like the Due or Zero. A 1024 bin FFT is perfectly practical.

But then you never said what frequency discrimination you need so how is anyone supposed to know.

7

Could you clarify the concepts that are expressed in the video?

Thanks

8

sossio89:
Could you clarify the concepts that are expressed in the video

It would take a book or at least a chapter of a book. Which I have just written.

Now if you could be more concise that just "the concepts" maybe I could help. Otherwise buy the book or get the information on line.
Here is a start:-

9

I know the theory of FFT.

I don't know the limits of Arduino in this application.

10

sossio89:
Could you clarify the concepts that are expressed in the video?

It's an example of the FFT capabilities of the Teensy3.1/SGTL5000 audio chip combination. As the Teensy is a lot faster than an UNO it can supposedly handle up to 512 bins with 43Hz step per bin so you could get a lot finer detail.

11

If I use a sample rate of 9000 hz and an FFT size of 512 bins. This means a signal from 0 to 4500 hz can be analyzed. Each FFT result bin will represent about 17 hz of frequencies (calculated by taking sample rate divided by FFT size).

It's right?

12

sossio89:
I know the theory of FFT.

But then why are you asking:-

If I use a sample rate of 9000 hz and an FFT size of 512 bins. This means a signal from 0 to 4500 hz can be analyzed. Each FFT result bin will represent about 17 hz of frequencies (calculated by taking sample rate divided by FFT size).

It's right?

No

  1. The number of bins used in the calculation has to be real data not the combined real and imaginary.
  2. You need two samples to recognise a frequency as at a sample rate of 9000 Hz the highest frequency you can detect is 4500Hz.

The Arduino Due and Zero as well as the Teensy are capable of doing this.

13

If I have this frequency to detect the failure:

294 Hz

205 Hz

20 Hz

273 Hz

Can i tell to Arduino (DUE or Zero)" found this frequencies and if their level is above a limit call me!"?

14

Can i tell to Arduino (DUE or Zero)" found this frequencies and if their level is above a limit call me!"?

Yes.

I would sample a buffer full of data and then run a band pass filter through it, one for each frequency you want to detect. The time is not important here, you could probably get a pass of all the frequencies about once every two seconds. A four pole recursive band pass filter should do it.
There is no need for an FFT.

Use the Direct Form I from this page:- Digital filter - Wikipedia

This might help you design it:- https://www-users.cs.york.ac.uk/~fisher/mkfilter/

15

@Grumpy_Mike: The filter page seemed interesting, but it does not work for me. I get this message in attempting to "design" a Hilbert Transform filter:

Bootleg
The page you have just come from is a bootleg copy of Tony Fisher's mkfilter page. It is out of date, unauthorized, and does not work. The real page is at:

http://www-users.cs.york.ac.uk/~fisher/mkfilter

16

OK that is the same as I posted, only found it through Google so I guess they are feeling miffed after the bad publicity surrounding their tax deal. :slight_smile:

17

Someone has a long memory! Tony Fisher evidently went to his reward in the year 2000.

18

No one's every truly dead on the internet tho ...

19

Now that is an interesting observation!
Not what I was taught as a child to imagine about the afterlife, tho...

20

Well, the afterlife, that's all superstition anyways, like Stevie Wonder sings "When you believe in things that you don't understand, Then you suffer, Superstition ain't the way.

Or as Mitsubishi advertised "Technically, Anything’s Possible".

Or as Nissan used to advertise "Oh, it's you Bob!"

Or my favorite, found on a meaning of names card in a Christian book shop (or similar) next to a children's clothing store we stopped in one time, and which I never let my wife forget, "Robert: Winner over all". Wish I'd managed to hang on to that one!